Machine Learning Models for Style Transfer in Sculpture

The application of machine learning to the field of sculpture has revitalised the boundaries of artistic invention and online production. The paper is about the high-order models of style transfer in sculpture which is the process of transferring visual characteristics of art in one type of sculpture to another not compromising on the structural integrity. Three models were employed including CycleGAN, Neural Style Transfer (NST) and Diffusion-based Generative Models. CycleGAN therefore facilitated unpaired data translation and hence was able to translate the material textures and sculptural patterns with an accuracy of style of 89.4. Neural Style Transfer had fine-grained aesthetic embedding at a texture fidelity of 85.7 percent on 3D surface renderings. Meanwhile, Diffusion Models were discovered to be more encouraging in spatial consistency and volumetric detailing with structural consistency of 92.1, superior to the traditional CNN-based techniques. The data were the digital scans of the classical and modern sculptures, the data were processed with the assistance of 3D voxel and mesh-based images. Experimentation has shown that a hybridization of these models is more expressive artistically and more efficient in calculations in the migration of sculptural style. The findings show that the latest AI-driven applications can not only imitate but also produce sculptural beauty and bridge the void between digital art and solid production with the assistance of 3D printing and CNC production. The piece is somehow related to the field of computational art in that it provides a possible paradigm of machine learning-like transfer to sculpture

Calligraphy · Sculpture · Texture synthesis · Visualization · Aesthetic Perception and Analysis · Generative Adversarial Networks and Image Synthesis · Music Technology and Sound Studies